Research Progress on Foveated Rendering Technology for Visual Comfort Optimization in Virtual Reality (Invited)

Significance As virtual reality display resolutions advance toward retinal-level standards, traditional graphics rendering pipelines face immense challenges regarding data throughput and computational load. Foveated rendering technology has emerged as a core method for balancing rendering efficiency and imaging quality by leveraging the non-uniform visual sensitivity of the human visual system (HVS). However, early research predominantly focused on static geometric simplification and lacked exploration of physiological discomforts such as visual fatigue and cybersickness during dynamic interactions, making it difficult to meet the demands of high immersion. As virtual reality has evolved from concept validation to commercial applications across healthcare, military, manufacturing, and entertainment domains, achieving visual comfort has become a critical challenge in transitioning VR systems from "usable" to "user-friendly." This paradigm shift requires addressing fundamental physiological conflicts such as the vergence-accommodation conflict (VAC) and visual-vestibular mismatches, which cause discomfort during prolonged use. Progress From 2010 to 2025, foveated rendering technology has evolved through distinct technical phases, moving from basic spatial sampling optimization toward comprehensive visual comfort enhancement. We classify these developments into five major categories. In spatial domain sampling approaches, early techniques like Guenter's multi-resolution nested framework established the foundation for geometric simplification based on visual acuity fall-off models. Later advances by Friston introduced perceptual rasterization that generates images with continuously-varying pixel density aligned with gaze position, eliminating artifacts from discrete layer transitions. Signal domain modulation methods leverage visual masking effects and contrast sensitivity functions to dynamically adjust rendering fidelity; Tursun's luminance-contrast-aware framework demonstrated that high-contrast regions can tolerate greater quality reduction without perceptible artifacts. The integration of cognitive awareness into rendering pipelines represents a significant paradigm shift, as shown by Krajancich's attention-aware model that quantifies how cognitive load can elevate peripheral visual thresholds by approximately 4 × during high-focus tasks. Most recently, neural rendering approaches have emerged, with Shi's scene-aware foveated neural radiance fields (SaF-NeRF) and Fan's Fov-GS for dynamic scenes using 3D Gaussian Splatting demonstrating how deep learning can reconstruct perceptually plausible details from extremely sparse samples. In visual comfort optimization, research has advanced from basic resolution reduction to physiological cue reconstruction, with Liu's perception-driven hybrid foveated depth-of-field framework addressing VAC through combined longitudinal chromatic aberration simulation and dynamic pupil response modeling. Conclusions and Prospects This review systematically summarizes the evolution of foveated rendering technology oriented toward visual comfort optimization from 2010 to 2025, revealing a fundamental paradigm shift from computational efficiency prioritization to perceptual quality management. Current research demonstrates that visual comfort cannot be achieved by resolution reduction alone but requires comprehensive modeling of physiological optical properties, cognitive attention mechanisms, and temporal stability. Despite significant progress, challenges remain in personalizing rendering parameters for individual visual characteristics, maintaining temporal coherence during rapid eye movements, and reducing computational latency for wireless VR applications. Future research should focus on developing lightweight neural rendering architectures with microsecond-level inference capabilities, exploring closed-loop adaptive calibration systems that incorporate physiological feedback signals (such as blink patterns, pupil dynamics, and vestibulo-ocular reflex metrics), and establishing standardized comfort assessment frameworks that bridge subjective experience with objective physiological measurements. These advancements will be crucial for next-generation virtual reality systems that deliver not only technological sophistication but also extended-wear comfort essential for mainstream adoption across professional and consumer applications.